shortestPath: Shortest Path Analysis
In GOstats: Tools for manipulating GO and microarrays

Description Usage Arguments Details Value Author(s) References See Also Examples

The shortest path analysis was proposed by Zhou et. al. The basic computation is to find the shortest path in a supplied graph between two Entrez Gene IDs. Zhou et al claim that other genes annotated along that path are likely to have the same GO annotation as the two end points.

1	shortestPath(g, GOnode, mapfun=NULL, chip=NULL)

`g`	An instance of the `graph` class.
`GOnode`	A length one character vector specifying the GO node of interest.
`mapfun`	A function taking a character vector of GO IDs as its only argument and returning a list of character vectors of Enterz Gene IDs annotated at each corresponding GO ID. The function should behave similarly to `mget(x, go2egmap, ifnotfound=NA)`, that is, `NA` should be returned if a specified GO ID has no Entrez ID mappings. See details for the interaction of `mapfun` and `chip`.
`chip`	The name of a DB-based annotation data package (the name will end in ".db"). This package will be used to generate an Entrez ID to GO ID mapping instead of `mapfun`.

The algorithm implemented here is quite simple. All Entrez Gene identifiers that are annotated at the GO node of interest are obtained. Those that are found as nodes in the graph are retained and used for the computation. For every pair of nodes at the GO term the shortest path between them is computed using sp.between from the RBGL package.

There is a presumption that the graph is undirected. This restriction could probably be lifted if there was some reason for it - a patch would be gratefully accepted.

The mapping of GO node to Entrez ID is achieved in one of three ways:

If mapfun is provided, it will be used to perform the needed lookups. In this case, chip will be ignored.
If chip is provided and mapfun=NULL, then the needed lookups will be done based on the GO to Entrez mappings encapsulated in the specified annotation data package. This is the recommended usage.
If mapfun and chip are NULL or missing, then the function will attempt to load the GO package (the environment-based package, distinct from GO.db). This package contains a legacy environment mapping GO IDs to Entrez IDs. If the GO package is not available, an error will be raised. Omitting both mapfun and chip is not recommended as it is not compatible with the DB-based annotation data packages.

The return values is a list with the following components:

`shortestpaths`	A list of the ouput from `sp.between`. The names are the names of the nodes used as the two endpoints
`nodesUsed`	A vector of the Entrez Gene IDs that were both found at the GO term of interest and were nodes in the supplied graph, `g`. These were used to compute the shortest paths.
`nodesNotUsed`	A vector of Entrez Gene IDs that were annotated at the GO term, but were not found in the graph `g`.

R. Gentleman

Transitive functional annotation by shortest-path analysis of gene expression data, by X. Zhou and M-C J. Kao and W. H. Wong, PNAS, 2002

sp.between

library("hgu95av2.db")
library("RBGL")

set.seed(321)
uniqun <- function(x) unique(unlist(x))

goid <- "GO:0005778"
egIds <- uniqun(mget(uniqun(hgu95av2GO2PROBE[[goid]]),
                            hgu95av2ENTREZID))

v1 <- randomGraph(egIds, 1:10, .3, weights=FALSE)
## Since v1 is random, it might be disconnected and we need a
## connected graph to guarantee the existence of a path.
c1 <- connComp(v1)
largestComp <- c1[[which.max(sapply(c1, length))]]
v2 <- subGraph(largestComp, v1)

a1 <- shortestPath(v2, goid, chip="hgu95av2.db")